Shugaban Kamfanin NVIDIA Jensen Huang ya gabatar da katunan bidiyo na wasan kwaikwayo na zamani da aka dade ana jira daga kicin dinsa. Kamar yadda aka sa ran, an sanar da tsofaffin mafita a yau: GeForce RTX 3090, GeForce RTX 3080 da GeForce RTX 3070. An gina katunan bidiyo akan Ampere generation GPUs da aka yi ta amfani da fasaha na 8nm na Samsung, yayin da aka samar da magabata na Turing ta amfani da fasaha na 12nm TSMC.
GeForce RTX 3090
Canji zuwa sabon tsarin fasaha mai zurfi ya ba da damar sanya ƙarin transistor a guntu, saboda abin da adadin tubalan aiki ya ƙaru sosai. Abin baƙin ciki shine, Jensen Huang bai fayyace daidaitawar GPU ba, kuma kawai suna suna sigogin tsarin ƙwaƙwalwar ajiya a cikin halayen. GeForce RTX 3090 yana sanye da 24 GB na ƙwaƙwalwar GDDR6X tare da ingantaccen saurin agogo na 19,5 GHz. Tare da bas mai faɗi 384-bit, wannan yana ba da bandwidth subsystem bandwidth na 936 GB/s. GeForce RTX 2080 Ti, tuna, yana da 11 GB na ƙwaƙwalwar GDDR6 kawai tare da ƙarancin bandwidth sau ɗaya da rabi.
Babban aikin GPU na sabon katin bidiyo shine 36 TFLOPS "tare da ginin hoto na gargajiya ta amfani da shading (shaders)" (Shader-TFLOPS). Aiki lokacin aiki tare da gano ray an bayyana shi a 69 TFLOPS (RT-TFLOPS), kuma aiki a cikin ayyuka ta amfani da muryoyin tensor ya kai 285 TFLOPS (Tensor-TFLOPS).
A cewar NVIDIA, an tsara wannan katin bidiyo don yin wasa a ƙudurin 8K, wanda ke da ikon samar da tsayayyen 60 FPS. Hakanan an lura cewa wannan "dabba" yana da kusan 50% sauri a ƙudurin 4K fiye da Titan RTX.
GeForce RTX 3080
Shugaban NVIDIA ya kira katin bidiyo na GeForce RTX 3080 sabon flagship, yana ƙetare tambayar menene GeForce RTX 3090 yake a wannan yanayin.
Daga cikin halayen, an lura da kasancewar 10 GB na ƙwaƙwalwar GDDR6X tare da ingantaccen mitar 19 GHz. Ana haɗa ƙwaƙwalwar ajiya ta hanyar bas 320-bit, wanda a ƙarshe yana ba da kayan aiki na 760 GB/s. Bari mu tuna cewa GeForce RTX 2080 na baya yana da 8 GB na ƙwaƙwalwar GDDR6 kawai tare da bandwidth na 448 GB/s.
Ayyukan GeForce RTX 3080 tare da fassarar gargajiya ta amfani da shaders shine 30 TFLOPS (Shader-TFLOPS). Lokacin aiwatar da binciken ray, katin bidiyo yana ba da 58 TFLOPS (RT-TFLOPS), kuma a cikin aiki akan nau'ikan muryoyin tensor yana ba da har zuwa 238 TFLOPS (Tensor-TFLOPS). A cewar shugaban NVIDIA, katin bidiyo na GeForce RTX 3080 ya ninka sau biyu kamar wanda ya gabace shi, GeForce RTX 2080.
GeForce RTX 3070
Katin mafi sauƙi na ukun da aka gabatar, GeForce RTX 3070, kamar wanda ya riga shi GeForce RTX 2070, yana da 8 GB na ƙwaƙwalwar GDDR6.
Matsayin aikin na'urar haɓaka zane-zane na GeForce RTX 3070 shine 20 TFLOPS tare da fassarar gargajiya tare da shaders (Shader-TFLOPS). Aiki lokacin aiki tare da gano ray an bayyana shi a 40 TFLOPS (RT-TFLOPS), kuma aiki a cikin ayyuka tare da muryoyin tensor ya kai 163 TFLOPS (Tensor-TFLOPS). NVIDIA ta yi iƙirarin cewa sabon GeForce RTX 3070 ya fi inganci fiye da na baya - GeForce RTX 2080 Ti.
Na farko na sababbin samfuran zai zama GeForce RTX 3080. Wannan zai faru a wannan watan - Satumba 17. Farashin zai kasance a matakin magabata - $ 699 (a cikin Rasha - 63 rubles). Mafi araha GeForce RTX 500 zai bayyana akan shelves wani lokaci a cikin Oktoba, kuma farashinsa zai zama $3070 (499 rubles) - kuma, a matakin wanda ya gabace shi. A ƙarshe, GeForce RTX 45 za ta ci gaba da siyarwa a ranar 500 ga Satumba, nan da nan ana samun su daga duka NVIDIA da abokan haɗin gwiwa, akan farashin ... $ 3090 (24 rubles). Babu wani karin farashi a nan.
source:
source: 3dnews.ru